CIRED
21st International Conference on Electricity Distribution
Frankfurt, 6-9 June 2011
Paper 0408
AUTOMATED ELECTRICAL ENERGY ANALYSIS FOR
DOMESTIC CONSUMERS BASED ON SMART METERS
Christian ELBE
Graz University of Technology - Austria
christian.elbe@tugraz.at
ABSTRACT
The European Union plans a roll out of smart meters to at
least 80 % of all households. Consumers should benefit
through more frequent billing periods and they should be
able to regulate their energy consumption through the total
power consumption provided by smart meters. Due to the
large number and the individual usage of electrical devices,
consumers have to spend a lot of time and effort to
distinguish between the influences regarding the total
power consumption of each electric device. In an average
Austrian household there are 9 major electrical devices that
have a total share of 72 % of the total power consumption.
To split up the bigger part of the total power consumption
additional measurement equipment is necessary. In order to
benefit from synergies without causing high costs for the
purchase and installation of measuring devices (with high
investment and operating costs) and without neglecting
ease of operation it is proposed that a mini PC, which
gathers and analyses the data from smart meters, has to
apply the method of non-intrusive load monitoring and
provide consumers with valuable additional information
about their power consumption of selected electrical
devices.
INTRODUCTION
The European Union set itself the target to reduce the
primary energy consumption by 20 % by 2020 compared to
projections for 2020. To achieve these reductions the EU's
energy efficiency policy relies on five pillars [1]. One of
these pillars is the general policy framework to which the
Directive 2006/32/EC belongs. Besides other measures
member states shall ensure that final customers are provided
with “…individual meters that accurately reflect the final
customer's actual energy consumption and that provide
information on actual time of use.” Furthermore “…billing
on the basis of actual consumption shall be performed
frequently enough to enable customers to regulate their own
energy consumption” [2].
Through the “third energy package” of the EU, member
states shall ensure “…the implementation of intelligent
metering systems that shall assist the active participation of
consumers in the electricity supply market”. Until 3rd
September 2012 an assessment “…of all the long-term costs
and benefits to the market and the individual consumer or
Paper No 0408
Ernst SCHMAUTZER
Graz University of Technology - Austria
schmautzer@tugraz.at
which form of intelligent metering is economically
reasonable and cost-effective and which timeframe is
feasible for their distribution…“ has to be accomplished by
the member states. A positive assessment leads to a roll out
of smart meters to at least 80 % of consumers [3].
In most recently published studies and smart grid demo
projects the main focus is on how electric utilities can use
this new technology to reduce costs and get more
information like loading or voltage characteristics about
low-voltage grids.
However, all approaches in reducing the energy
consumption in households are based on informing
consumers about the total power consumption of a
selectable period. Due to the increasing number of electrical
devices in households, it is not possible for customers to
recognize easily which electric devices have a significant
share of the total power consumption. Impacts of changes in
the user behaviour as well as realized energy efficiency
actions are not traceable for them through observing the
total power consumption. Due to these reasons customers
are not able to identify and monitor the influencing factors
on their electrical energy demand or even separate the total
power consumption by major electrical devices.
AUTOMATED ENERGY ANALYSIS
Providing consumers with information about their electrical
energy consumption during a selectable period like the
European Commission is planning is in any case an
improvement regarding the status quo. However,
information about the total power consumption doesn’t give
consumers the opportunity to differ between the
consumption of single electrical devices. Several currently
sold smart meters provide the possibility to log load curves
with a home PC. Despite this, manually observing such load
curves is a time consuming analysis assuming special
expertise about the behaviour of different electrical
appliances as well as a diary of the usage of single electrical
appliances. Without a handwritten log about the usage of
major appliances it is infeasible to reveal which devices
have contributed to the monitored total power consumption.
Besides this a high number of different electrical devices
switched on at the same time make such an analysis more
complicated. So an automated analysis of this data would be
the only possible and practicable solution for the consumer.
For measuring the electrical power consumption of major
electrical devices with traditional energy-monitoring
1/4
CIRED
21st International Conference on Electricity Distribution
Frankfurt, 6-9 June 2011
Paper 0408
devices, meters for each load are required which cause high
investment and operating costs. In order to benefit from
synergies without causing high costs for the purchase and
installation of measuring devices, in this paper a concept is
described for gathering data from smart meters to apply the
method of non-intrusive load-monitoring (NILM).
Concept
A mini PC which costs below 100 EUR having a rated
power of about 5 W is connected to a smart meter and
records active and reactive power as well as the voltage
which allows normalizing resistive loads with a high
resolution in time. The recorded data is analyzed frequently
by an algorithm. Recurring patterns from electrical
appliances with high loads are detected and the power
consumptions of these devices are calculated. The hourly
power consumption of all detectable electrical devices as
well as further statistical data about the usage and the
detailed electrical costs are provided in diagrams and tables
(see figure 1). Measures like the replacement of an electrical
device with a significant share to the total power
consumption or additional energy consuming electrical
devices are traceable for consumers through automated
analysis via visualization. Energy-hungry electrical devices
can be identified easily and consumers can find out the
standby power consumption. Besides this, influences of the
user behaviour on the energy consumption of electrical
devices are analysed. Each household gets individual
information how to save energy and which are the most
efficient measures. For a deeper analysis parts of the
monitored data can be provided to energy savings
consultants who can give more detailed information about
further measures and energy savings advices. Through the
development of intelligent electrical devices which are
equipped with communication chips the optimizations
targets of the automated analysis can be varied between load
shift and peak clipping to reduce GHG-emissions on its
maximum level.
Furthermore the power consumption of selected electrical
devices is automatically rated by several inputs of
parameters such as energy efficiency grade or size. So
consumers can compare their devices with standard devices
and the best available technology with the lowest energy
consumption on the market. Through an energy savings
evaluation tool the amortization time for different electrical
appliances as well as provided measures can be calculated.
The software for the algorithms is automatically updated
over the internet. All the monitored data is only saved on
the mini PC which is fixed in the distribution box. Merely
consumers have access through LAN or WLAN to this
device which is protected by encryption. The results of the
automated energy analysis are visually displayed for
customers.
Monitoring measures for saving energy
The total power consumption in households rises by time
through additional electrical appliances or replacement by
bigger devices like large widescreen plasma televisions.
Furthermore the total power consumption varies day by day
and each electrical device has an individual usage.
Influences of a single electrical device on the total power
consumption are not traceable for most consumers
nowadays. Even when measures to reduce the total electric
Figure 1: A mini PC which is connected to a smart meter detects and analyzes the power consumption of several
electrical devices and provides an interface for consumers.
Paper No 0408
2/4
CIRED
21st International Conference on Electricity Distribution
Frankfurt, 6-9 June 2011
Paper 0408
power consumption are realized, it is not sure that these
measures are reflected in the total power consumption. Due
to the reason that the total power consumption is dependent
on so many parameters such measures are just traceable
with special expertise. The total power consumption varies
over the year (blue line) and raises yearly (dashed black
line) as can be seen in figure 2. Besides this the total power
consumption varies day by day what leads to a baseline
conflict.
W in p.u.
1.5
+ W
1
- W
0.5
0
0
0,25
0,5
0,75
1
t in a
Figure 2: Variation of the average total power
consumption during a year
Through the automated energy analysis of the load curve a
couple of measures such as saving electrical energy for
heating of water in electric heaters or washing machines as
well as additional consumption can be monitored and
analysed. For example the frequency of the usage of
electrical devices such as washing machines and the
variation of the time for heating up water give information
about how energy efficient the users operate electrical
devices. A ranking allows consumers to compare their
power consumption of a single electrical device with other
consumers’ devices.
Another example is the observation of the frequency of the
cycle time of the heater of a stove. When consumers tend to
use the maximum heating level of a stove very often it can
be sure that they do not use lids on their pots which can be
detected by the automated energy analysis. So, specific
advices can be determined.
energy consumption below 15% [5].
Another recent approach by [6] detected major electrical
appliances such as stove, oven, geyser, microwave and
kettle by step changes in total active power with a maximum
error of 9 % in energy consumption. Considering the power
consumption of each phase of a three-phase power system
separately and observe the reactive power consumption as
well as maximizing the resolution of the measurement data
to at least 0.1 seconds would also raise the correct detection
rate of appliances. Even an error of 9 % would be enough to
give consumers a sufficient overview of their share in total
electrical power consumption and provide them information
about how much money they spend for each major electrical
appliance and the rest (standby,…).
Due to the increasing number of electrical loads and varying
electrical power consumption as well as load profile
overlapping of electrical appliances it is not possible to
detect all devices with an adequate minimal error in energy
consumption. Electrical devices with recurring patterns and
high loads are easier to detect.
Total electrical energy share
Just a couple of electrical appliances are responsible for the
bigger part of the total electrical power consumption. A
study carried out by the Statistics Austria come to the
conclusion that 72 % of the total electrical energy
consumption of an average Austrian household is caused by
9 different major electrical devices as well as standby [7]
(see Figure 3). All of these devices have recurring patterns
with high loads and can be detected by available NILM
algorithms. Dependent on the individual household and the
existing equipment of electrical appliances this share can
vary and consequently also the detection rate of the NILM
algorithms. Nevertheless it can’t be neglected that
households would benefit through additional information of
an automated analysis about their individual electric energy
consumption even with a lower share.
Non-intrusive Load Monitoring
Hart [4] proposed a NILM system that is based on changes
of active and reactive power of electrical appliances. Hart’s
concept was to represent each electrical appliance by a
finite state machine which handles discrete power
consumption changes for gathering states.
Since 1992 a lot of papers have been proposed new
algorithms such as harmonic pattern recognition, time
frequency characterization, using steady-state and turn-on
transient energy based on artificial neural networks, hidden
markov models or integer programming. While it is nearly
impossible to successfully detect all appliances with power
consumption below 100 W, the recognition of devices that
have a recurring load profile and high loads such as water
heaters or refrigerators works with an error in evaluating the
Paper No 0408
Figure 3: Average total electrical power consumption of
an Austrian household [7]
Gathering Measuring Data by Smart Meters
Smart meters measure voltages and currents with a sampling
rate of typically 12 to 15 kHz and the active or
instantaneous power is processed internally with a high
resolution of a couple of milliseconds for all 3 phases.
3/4
CIRED
21st International Conference on Electricity Distribution
Frankfurt, 6-9 June 2011
Paper 0408
These data could but currently can’t be used to apply the
method of nonintrusive load monitoring.
For getting good results and minimum detection errors with
NILM algorithms, a load curve with a resolution of at least
one second is necessary. The higher the resolution in time of
active and reactive power the easier overlaps of load curves
from different appliances can be distinguished. Recently
published NILM algorithms analyse harmonics as well as
turn on transient energy. These algorithms typically process
reactive and active power in milliseconds and current
harmonics with several kHz. Though such algorithms can
raise the detection rate and give the opportunity to detect
further electrical appliances, analysing reactive and active
power with a resolution of about one second, gives
maximum detection errors of 9% in energy consumption
[6]. However, to gather the measuring data for NILM
algorithms from smart meters the minimum resolution of
reactive and active power as well as the voltage for all three
phases should go below one second.
Interfaces of smart meters
Currently sold smart meters are equipped with pulse
indication, a multipurpose expansion port, a wire-connected
PLC interface and an optical service interface. Through
expansions for the multipurpose interface radio ZigBee
connections and wireless M-Bus connections can be
established. Among other things the optical service interface
provides the possibility to read out standardized parameters
(IEC 62056-21) such as voltage, electrical power
consumption per phase or 15-minute-averages of active
power. These interfaces are predominantly used for in-home
displays and connections to other digital meters. Up to now
a few manufacturers provide in collaboration with software
developers off-the-shelf solutions for consumers to display
the current total power consumption on PCs every second.
Table 1 shows an overview of available interfaces of
currently sold smart meters in Europe.
MAX.
RESOLUTION
P/Q
PER PHASE
MAX.
RESOLUTION
VOLTAGE
Service-Interface
(IEC 62056-21)
1-5 seconds
1-5 seconds
pulse indication
up to 60.000
pulses per kWh
or kvar
n.a.
Multipurpose
Expansion Port
P: 2-4 seconds
n.a.
INTERFACE
inhouse:
inhouse:
1 second,
wire-connected
n.a.,
external:
(PLC)
external:
15-minute15 minutes
average
Table 1: Overview of available Interfaces of Smart
meters different manufacturers
Paper No 0408
So far, it is not possible to read out active and reactive
power as well as the voltage per phase every second which
is necessary for an automated energy analysis. Although the
internal data is processed in milliseconds there is no
possibility to read out this data with a high resolution
including a timestamp.
However, to unfold the full potential for saving energy with
the help of smart meters, manufacturers should be aware of
these lacks and provide as soon as possible though smart
meters.
CONCLUSION
The total power consumption doesn’t allow distinguishing
between the power consumption of different electrical
appliances. Just a daily log of the usage of electrical
appliances allows consumers with special expertise to
retrace the variations in the total power consumption. The
efforts from the EU to reduce GHG emissions through smart
meters which provide the total power consumption can only
be done with a necessary scale by consumers with special
expertise doing great efforts.
With a mini PC, which could also be integrated in the next
generation of smart meters, an automated energy analysis
can be accomplished that gives consumers detailed
information about their individual power consumption. Such
an automated energy analysis which gives specific advices
for saving energy allows the maximum savings in the total
power consumption. Furthermore appropriate optimization
targets such as energy saving, load shifting or peak clipping
combined with intelligent electrical devices can reduce the
GHG emissions to its minimum.
REFERENCES
[1] European Commission, 2008, “Energy efficiency:
delivering the 20% target”, Communication from the
Commission, COM(2008) 772 final, 5-6.
[2] European Parliament, 2006, “Directive 2006/32/EC”,
Article 13.
[3] European Parliament, 2009, “Directive 2009/72/EC”,
Annex I.
[4] G.W. Hart: “Nonintrusive appliance load monitoring”,
Proceedings of the IEEE, Vol. 80, No.12, pp.1870–
1891(1992).
[5] L. Farinaccio, R. Zmeureanu, 1999, “Using a pattern
recognition approach to disaggregate the total
electricity consumption in a house into the major enduses”, Energy and Buildings 30 (1999), 245-259.
[6] Arend J. Bijker, Xiaohua Xia, Jiangfeng Zhang,
“Active Power Residential Non-intrusive Appliance
Load Monitoring System”, IEEE AFRICON 2009.
[7] A. Wegscheider-Pichler, 2009, Strom- und
Gastagebuch 2008, Statistics Austria, Vienna, Austria.
4/4